LWiAI Podcast #235 - Sonnet 4.6, Deep-thinking tokens, Anthropic vs Pentagon

Read the full articleLWiAI Podcast #235 - Sonnet 4.6, Deep-thinking tokens, Anthropic vs Pentagon on Last Week in AI

What Happened

Anthropic releases Sonnet 4.6, Google Rolls Out Gemini 3.1 Pro, Anthropic CEO Amodei says Pentagon’s threats ‘do not change our position’ on AI

Our Take

anthropic dropping sonnet 4.6 and we're hearing the same old noise about the pentagon drama. they're talking about deep-thinking tokens and us military threats, and it just confirms what we already knew: this isn't about clever algorithms; it's about who controls the access keys to the next generation of compute.

google rolling out gemini 3.1 pro right alongside it just means the arms race is getting exponentially faster. the tech specs matter less than the fact that two major players are fighting over the infrastructure and the rules of engagement.

this isn't research; it's a power play disguised as a model release. the focus should be on the actual compute costs and where the guardrails are, not the rhetoric.

What To Do

map out which vendor dependencies create the most security bottlenecks. impact:high

Builder's Brief

Who

teams building dual-use or defense-adjacent AI applications on Anthropic APIs

What changes

Anthropic's acceptable use policy is under active political pressure; compliance posture for sensitive sectors may shift without notice

When

months

Watch for

Anthropic publishing a revised acceptable use policy or losing a named enterprise contract publicly

What Skeptics Say

Amodei's public Pentagon defiance is undermined by the fact that Claude is actively used in military targeting decisions — the stated policy and the actual usage are already in contradiction.

Cited By

Last Week in AI LWiAI Podcast #235 - Sonnet 4.6, Deep-thinking tokens, Anthropic vs Pentagon